http://www.perlmonks.org?node_id=151546


in reply to Unrolling recursion

One general technique is to use a stack:
seed the stack with the first instance while (stack has entries) pop off top entry from stack process it take any potential new work as a result of processing, put it back on the stack
And depth/breadth first recursion can be adjusted by shifting instead of popping at the top of the loop (or when inserting new work).

Of course this only replaces the subroutine stack with your own, but it can be quicker and save memory. (Sometimes it's a lot more obvious too!)

Replies are listed 'Best First'.
Re: Re: Unrolling recursion
by clintp (Curate) on Mar 14, 2002 at 15:15 UTC
    From a /msg request for dragonchild, here's pseudocode to do a directory traversal without recursion.
    # Traverse the filesystem push to stack / while currentdirectory = pop stack open the currentdirectory foreach pathname in this directory if it's also a directory, push it (full path) on the stack close the current directory
    As opposed to a recursive solution:
    call traverse(/) function traverse (currentdirectory) open currentdirectory foreach pathname in this directory if it's also a directory, traverse(pathname) close currentdirectory
      For those who don't grok pseudocode here's a perl translation for directory traversal without recursion
      my @stack = "."; while(my $cd = pop @stack) { opendir(my $dh, $cd) or die("bad dir - $!"); foreach (readdir($dh)) { next if /^\./; print $_.$/; push @stack, "$cd/$_" if -d "$cd/$_"; } }
      and with recursion
      &traverse("."); sub traverse { my $dir = shift; opendir(my $dh, $dir) or die("bad dir - $!"); foreach (readdir($dh)) { next if /^\./; print $_.$/; &traverse("$dir/$_") if -d "$dir/$_"; } }
      That's a pretty neat 'trick' which I guess might be how optimising compilers implement some forms of recursion (or something).
      HTH

      broquaint